Enable reporting unknown types #333
Conversation
ItsDrike
added
t: enhancement
ItsDrike
force-pushed
the
enable-unknown-type-reporting
branch
from
6cbd87e
to
e434651
Compare
Since we have the reportUnknownParameterType for basedpyright enabled, we can't leave `*args` or `**kwargs` unannotated anyways, so we might as well enforce it properly.
| @@ -110,7 +110,7 @@ | |||
| autodoc_member_order = "bysource" | |||
|
|
|||
| # Default options for all autodoc directives | |||
| autodoc_default_options = { | |||
| autodoc_default_options: dict[str, Any] = { | |||
There was a problem hiding this comment.
I don't see how these kind of type annotations are better
There was a problem hiding this comment.
Yeah, the thing is, pyright would otherwise mark this as dict[str, Unknown] and with Unknown being banned, if a wildcard type is indeed what we want, it must be declared explicitly like this. That's also one of the reasons why these rules might be a bit too strict.
| @@ -153,7 +153,7 @@ def run(self) -> list[AttributeTablePlaceholder]: | |||
| def build_lookup_table(env: BuildEnvironment) -> dict[str, list[str]]: | |||
| # Given an environment, load up a lookup table of | |||
| # full-class-name: objects | |||
| result = {} | |||
| result: dict[str, list[str]] = {} | |||
There was a problem hiding this comment.
But for these ones it's interesting to have type checking
There was a problem hiding this comment.
I agree, in some cases, being forced to explicitly declare the annotations can be very helpful to the type checker, it is nice that this would now be a requirement, where types can't be easily inferred.
| @@ -66,7 +66,7 @@ def __init__(self, exc: httpx.HTTPStatusError): | |||
| @property | |||
| def msg(self) -> str: | |||
| """Produce a message for this error.""" | |||
| msg_parts = [] | |||
| msg_parts: list[str] = [] | |||
There was a problem hiding this comment.
I already have things like that all over the place in more-packets
| @@ -477,7 +479,7 @@ def test_nbt_bigfile(): | |||
| data = bytes.fromhex(data) | |||
| buffer = Buffer(data) | |||
|
|
|||
| expected_object = { # Name ! Level | |||
| expected_object: FromObjectType = { # Name ! Level | |||
There was a problem hiding this comment.
This I find useful
|
I'm not against these new rules as I feel they are not the most annoying of the bunch, but I don't really know if the extra effort is worth the while. |
|
The worst / most annoying stuff are recursive types, like those in |
|
Type checking is useful when it uses the computer's power to mitigate our forgetfulness, but when it's going the opposite way I find it annoying 😆 . |
This PR is an experiment, just to see how the code-base would look like if we enabled reporting unknown types for basedpyright.
It's more of a prompt for discussion on whether or not this is worth it, I do not intend to merge this, at least not before this discussion takes place.
The issue with these rules is that they can sometimes be incredibly annoying to satisfy, it often means having to use a bunch of type-casts or just ignoring them. However, in many cases, they can also be incredibly useful, as it's good to know when basedpyright can't recognize some type and needs some help in understanding it. Overall, it would definitely improve the type safety, but the question is whether it's even worth it.